CN114048600A - Digital twin-driven multi-model fusion industrial system anomaly detection method - Google Patents
Digital twin-driven multi-model fusion industrial system anomaly detection method Download PDFInfo
- Publication number
- CN114048600A CN114048600A CN202111320896.8A CN202111320896A CN114048600A CN 114048600 A CN114048600 A CN 114048600A CN 202111320896 A CN202111320896 A CN 202111320896A CN 114048600 A CN114048600 A CN 114048600A
- Authority
- CN
- China
- Prior art keywords
- model
- data
- digital twin
- abnormal
- driven
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a digital twin driven multi-model fusion industrial system anomaly detection method, which comprises the following steps: firstly, constructing a physical model and integrating the physical model into a unified multi-field system digital twin model; secondly, carrying out physical model simulation and iterative modification; thirdly, performing information modeling by using an edge calculation technology; fourthly, constructing a data-driven neural network intelligent detection model; fifthly, preprocessing the data obtained in the fourth step; introducing a threshold sorting method to predict abnormal features; seventhly, storing the data into a cloud platform database for extracting abnormal features, and constructing a final abnormal detection model; and tenthly, saving the corresponding abnormal data to a database as data for updating the optimization model. The invention integrates the data-driven intelligent deep learning detection model and the digital twin-driven real-time information model, overcomes the problems of low accuracy based on a physical model and poor adaptability of a data-driven method, and greatly improves the accuracy of a prediction result.
Description
Technical Field
The invention belongs to the field of intelligent manufacturing and industrial engineering, and particularly relates to a digital twin-driven multi-model fusion industrial system anomaly detection method.
Background
In the context of smart manufacturing, the functionality of industrial systems becomes more complex and larger, which makes the detection and maintenance of industrial systems more challenging. Accurate and efficient fault detection is crucial to the quality, safety and efficiency of industrial production. However, industrial data has the characteristics of large data volume, more noise interference, coupling of various data and the like, so that some existing detection schemes are low in efficiency and poor in real-time performance. Various industries seek new anomaly detection methods, and in recent years, intelligent anomaly detection methods based on technologies such as intelligent sensing information, big data analysis technology and cloud computing have attracted much attention.
Anomaly detection minimizes production lifecycle costs by eliminating potential hazards before machine failures occur. The anomaly detection of an industrial system mainly comprises continuous equipment monitoring by using a sensor and anomaly detection by using a detection model. The current industrial system anomaly detection technology mainly faces the following challenges: 1) it is difficult to efficiently process the large amount of data generated in an industrial system. Although the popularization of the internet of things technology enables real-time data collection of industrial systems, a huge barrier still exists in data collection and integration at present due to the difficulty in converting a large amount of data into meaningful information. 2) The operating environment and the working condition of the industrial system change in real time, the existing model always assumes that fault data which may appear in the future is the same as the existing data distribution, and the detection model is difficult to adapt to the distribution change of the data. 3) Industrial systems are complex and typically consist of multiple subsystems, devices, and components that are interdependent and interactive with each other. Different types of faults may occur simultaneously on different components or subsystems. The signals collected by the sensors are often noisy, and various fault features are compounded in a nonlinear form and are difficult to accurately detect. 4) The detection of the state of the industrial system involves a plurality of disciplines and information such as electricity, structure, energy, heat, mechanics and the like, and different disciplines are coupled in a complex manner to form the industrial system with complex structure and function. The system state needs to be comprehensively analyzed by combining knowledge of a plurality of disciplines. 5) It is impossible to perform abnormality detection by fully utilizing the existing history data.
Aiming at the problems of the existing industrial system detection technology, the invention introduces the technologies of digital twinning, edge calculation, deep learning and the like to construct a data-driven hybrid anomaly detection framework, and has important reference value for predictive maintenance and anomaly detection of the actual industrial system.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a digital twin-driven industrial system anomaly detection framework aiming at the defects in the prior art, the framework integrates a data-driven intelligent deep learning detection model and a digital twin-driven real-time information model, and the problems of low accuracy based on a physical model and poor adaptability of a data driving method are solved, so that the accuracy of a prediction result is greatly improved.
In order to solve the technical problems, the invention adopts the technical scheme that: a digital twin driven multi-model fusion industrial system anomaly detection method comprises the following steps:
firstly, constructing a physical model for subsystem levels and equipment based on a digital twin technology and combining software such as ANSYS, MWorks, Dymola, Simulinex and the like, and integrating the physical model into a unified multi-field system digital twin model;
performing physical model simulation and performing iterative modification and simulation on the multi-field system digital twin model to ensure the accuracy of the digital twin model until the error between a simulation result and an experiment result is small enough to obtain a high-fidelity digital twin model, and establishing a mapping interface to ensure the real-time update of the digital twin model;
performing information modeling on the relation between the physical entity and the digital twin body at an edge layer by using an edge computing technology to ensure real-time interaction and synchronization between the physical entity and the digital twin body;
collecting data generated by various intelligent sensors in the operation process of the industrial system, and constructing a data-driven neural network intelligent detection model based on past abnormal data of the industrial system;
taking the data obtained in the fourth step as a sample, performing noise reduction, instance selection, normalization, data cleaning and preprocessing on the data obtained in the fourth step on an edge layer by using edge calculation, and processing the problem of sample imbalance by using an XGboost algorithm;
step six, introducing a threshold sorting method (The Ranking and threshold method) to carry out a plurality of predictions on abnormal features, thereby solving The problem of mutual interference among a plurality of abnormal features possibly contained in The data and accurately diagnosing The abnormal features as The abnormality of a plurality of corresponding parts;
step seven, transmitting the data obtained in the step four to a cloud platform, carrying out model training, iterating and updating parameters, training a plurality of groups of data sets to obtain an abnormality detection model, wherein the abnormality detection model consists of a plurality of prediction models, the most significant parameters of the target parameters are determined by a Pearson correlation coefficient method, and the coefficients can be obtained by the following formula:
where n is the number in the data set and x, y are the parameter values and the data-driven anomaly detection model is updated using the partial historical data associated with these parameters as inputs to the model. Continuously learning new knowledge from new real-time data by using incremental learning, adjusting an algorithm and updating a model;
step eight, integrating the abnormal detection models of the subsystems and the equipment in the industrial system through a fusion method, and distributing different weights to form a system-level abnormal detection model;
step nine, based on a hybrid Algorithm fusion algorithm, combining a digital twin model and a data-driven neural network intelligent detection model, setting and inputting the digital twin model as a space state to construct a final abnormal detection model;
step ten, deploying the trained anomaly detection model on the edge layer to perform real-time anomaly detection; storing the prediction result in an edge layer, verifying the accuracy of the anomaly detection model by comparing the actual running state with the prediction result, evaluating the performance of the model, and determining whether the algorithm needs to be updated or retrained;
step eleven, setting an evaluation mechanism for the output of the abnormality detection model; since anomaly detection is a continuous process, the health of the device is difficult to determine from the values at a particular point in time. The situation that the abnormal condition appears temporarily and the normal range is returned quickly is eliminated, and the condition that the abnormal condition continuously exceeds the normal range is focused; determined by the following evaluation mechanism:
wherein P isi(Yi(t)) is the evaluation result of the detection model output, f is the input signal frequency, when Si(t) the system will make an early warning if it is less than a threshold set based on expertise and experience.
Step twelve, executing a corresponding maintenance plan according to the obtained abnormal detection result, and saving corresponding abnormal data to a database as data for updating the optimization model;
in the above method for detecting the abnormality of the digital twin-driven multi-model fusion industrial system, the specific process of the multi-field system digital twin model constructed on the subsystem level and the equipment in the step one is as follows:
step 101, collecting the operation condition and working condition data of an industrial system provided by an intelligent sensor as input parameters of a model;
102, performing modeling simulation on a control system, a material system, transmission equipment, a production device subsystem and equipment by using software such as ANSYS, MWorks, Dymola, SimlationX and the like;
103, optimizing the simulation model based on multi-field expert knowledge and empirical data of machinery, heat, mechanics and automation, and integrating models of all subsystems and equipment parts into a multi-field system digital twin model.
In the method for detecting the abnormality of the digital twin-driven multi-model fusion industrial system, the specific process of performing information modeling on the real-time interaction and synchronization between the physical entity and the digital twin body at the edge layer by using the edge computing technology in the third step is as follows:
step 301, establishing a geometric model for a physical entity at a cloud end for geometric description and kinematic description of a physical system;
step 302, constructing a functional model describing behaviors and functions of the digital twin to manage the functions and behaviors of the digital twin;
step 303, modeling the information of the connection between the physical entity and the digital twin at the edge layer.
The method for detecting the abnormality of the digital twin-driven multi-model fusion industrial system comprises the following steps of constructing a data-driven neural network intelligent detection model, wherein the specific process comprises the following steps:
step 401, collecting past abnormal data of the industrial system as training data of the model, simultaneously collecting data information of the industrial system obtained by the intelligent sensor under different working conditions, transmitting the data information through MQTT and participating in training to improve the generalization capability of the model;
step 402, constructing an abnormal feature extractor based on a deep neural network, and extracting abnormal features layer by layer through forward propagation;
step 403, mapping the data with different probability distributions in each domain to a high-dimensional space through metric learning, reducing the data distribution difference among different domains, and obtaining the data distribution difference loss;
and step 404, classifying the extracted system operation state features by using a classifier containing a full connection layer to obtain an abnormal detection result and a classification loss, and introducing a regularization term to improve the robustness of the model.
In the method for detecting the abnormality of the digital twin-driven multi-model fusion industrial system, the specific processes of transmitting the data to the cloud platform, training the model, iterating and updating the parameters in the seventh step are as follows:
step 701, collecting and transmitting real-time data to an edge layer through an intelligent sensor and data acquisition equipment, preprocessing the real-time data at the edge layer to reduce transmission pressure, training the preprocessed data on a cloud platform, and storing the data transmitted to the cloud platform in a cloud database;
step 702, determining the most feasible anomaly detection model based on historical data and real-time data through an automatic machine learning (AutoML) optimization algorithm and hyper-parameter selection;
and 703, learning new knowledge from the new real-time data by using incremental learning, adjusting an algorithm and updating an anomaly detection model.
In the above method for detecting an abnormality of a digital twin-driven multi-model fusion industrial system, the specific process of constructing a final abnormality detection model based on the hybrid algorithm in combination with the digital twin model and the data-driven neural network model in the ninth step is as follows:
step 901, setting a digital twin model as a space state and an operation condition of a system-level fusion model, and inputting a fusion algorithm based on internal values of a multi-physics simulation computing system;
step 902, obtaining a preliminary predicted value of each index of the system by a data-driven neural network intelligent detection model obtained by training;
and 903, integrating the digital twin model and the machine learning detection model by using a hybrid Algorithm fusion algorithm to obtain a more accurate detection result, judging whether each index reaches a threshold value, and iteratively updating the abnormal detection model.
Compared with the prior art, the invention has the following advantages:
1. the invention overcomes the problems that the modeling accuracy of the industrial system based on the physical model is low, and the modeling complexity is increased along with the increase of the complexity of the industrial system. The intelligent level of the detection method is greatly improved by applying machine learning.
2. The invention introduces the digital twinning and edge calculation technology to solve the problem that the traditional method is difficult to efficiently process a large amount of data generated in an industrial system.
3. Compared with the existing method, the method can better handle the condition that a plurality of exceptions exist in the industrial system at the same time. Aiming at the problems that different types of faults occur on different components or subsystems at the same time, the faults mutually affect each other and interaction is difficult to monitor, a threshold value sorting method is introduced, a multi-field model is constructed, various abnormal features are effectively separated, and detection precision is improved.
4. Compared with a general deep learning detection algorithm, the method has higher real-time performance. Under the drive of a digital twinning technology, the digital twinning body capable of reflecting the real state of a workshop is established on an information layer and is fused with an intelligent detection model established by a cloud end, so that the adaptability of the model to the change of an industrial system is greatly improved.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
FIG. 1 is a flow chart diagram of the anomaly detection method of the digital twin-driven multi-model fusion industrial system according to the present invention.
Detailed Description
As shown in FIG. 1, the invention relates to a digital twin-driven multi-model fusion industrial system anomaly detection method, which comprises the following steps:
firstly, constructing a physical model for subsystem levels and equipment based on a digital twin technology and combining software such as ANSYS, MWorks, Dymola, Simulinex and the like, and integrating the physical model into a unified multi-field system digital twin model;
in this embodiment, the specific process of the multi-domain system digital twin model established on the subsystem level and the device in the step one is as follows:
step 101, collecting the operation condition and working condition data of an industrial system provided by an intelligent sensor as input parameters of a model;
102, performing modeling simulation on a control system, a material system, transmission equipment, a production device subsystem and equipment by using software such as ANSYS, MWorks, Dymola, SimlationX and the like;
103, optimizing the simulation model based on multi-field expert knowledge and empirical data of machinery, heat, mechanics and automation, and integrating models of all subsystems and equipment parts into a multi-field system digital twin model.
Performing physical model simulation and performing iterative modification and simulation on the multi-field system digital twin model to ensure the accuracy of the digital twin model until the error between a simulation result and an experiment result is small enough to obtain a high-fidelity digital twin model, and establishing a mapping interface to ensure the real-time update of the digital twin model;
performing information modeling on the relation between the physical entity and the digital twin body at an edge layer by using an edge computing technology to ensure real-time interaction and synchronization between the physical entity and the digital twin body;
in this embodiment, the specific process of performing information modeling on the real-time interaction and synchronization between the physical entity and the digital twin at the edge layer by using the edge computing technology in the third step is as follows:
step 301, establishing a geometric model for a physical entity at a cloud end for geometric description and kinematic description of a physical system;
step 302, constructing a functional model describing behaviors and functions of the digital twin to manage the functions and behaviors of the digital twin;
step 303, modeling the information of the connection between the physical entity and the digital twin at the edge layer.
Collecting data generated by various intelligent sensors in the operation process of the industrial system, and constructing a data-driven neural network intelligent detection model based on past abnormal data of the industrial system;
in this embodiment, the step four of constructing the data-driven neural network intelligent detection model includes the specific processes of:
step 401, collecting past abnormal data of the industrial system as training data of the model, simultaneously collecting data information of the industrial system obtained by the intelligent sensor under different working conditions, transmitting the data information through MQTT and participating in training to improve the generalization capability of the model;
step 402, constructing an abnormal feature extractor based on a deep neural network, and extracting abnormal features layer by layer through forward propagation;
step 403, mapping the data with different probability distributions in each domain to a high-dimensional space through metric learning, reducing the data distribution difference among different domains, and obtaining the data distribution difference loss;
and step 404, classifying the extracted system operation state features by using a classifier containing a full connection layer to obtain an abnormal detection result and a classification loss, and introducing a regularization term to improve the robustness of the model.
Taking the data obtained in the fourth step as a sample, performing noise reduction, instance selection, normalization, data cleaning and preprocessing on the data obtained in the fourth step on an edge layer by using edge calculation, and processing the problem of sample imbalance by using an XGboost algorithm;
step six, introducing a threshold sorting method (The Ranking and threshold method) to carry out a plurality of predictions on abnormal features, thereby solving The problem of mutual interference among a plurality of abnormal features possibly contained in The data and accurately diagnosing The abnormal features as The abnormality of a plurality of corresponding parts;
step seven, transmitting the data obtained in the step four to a cloud platform, carrying out model training, iterating and updating parameters, training a plurality of groups of data sets to obtain an abnormality detection model, wherein the abnormality detection model consists of a plurality of prediction models, the most significant parameters of the target parameters are determined by a Pearson correlation coefficient method, and the coefficients can be obtained by the following formula:
where n is the number in the data set and x, y are the parameter values and the data-driven anomaly detection model is updated using the partial historical data associated with these parameters as inputs to the model. Continuously learning new knowledge from new real-time data by using incremental learning, adjusting an algorithm and updating a model;
in this embodiment, the specific process of transmitting data to the cloud platform, performing model training, iterating, and updating parameters in the seventh step is as follows:
step 701, collecting and transmitting real-time data to an edge layer through an intelligent sensor and data acquisition equipment, preprocessing the real-time data at the edge layer to reduce transmission pressure, training the preprocessed data on a cloud platform, and storing the data transmitted to the cloud platform in a cloud database;
step 702, determining the most feasible anomaly detection model based on historical data and real-time data through an automatic machine learning (AutoML) optimization algorithm and hyper-parameter selection;
and 703, learning new knowledge from the new real-time data by using incremental learning, adjusting an algorithm and updating an anomaly detection model.
Step eight, integrating the abnormal detection models of the subsystems and the equipment in the industrial system through a fusion method, and distributing different weights to form a system-level abnormal detection model;
step nine, based on a hybrid Algorithm fusion algorithm, combining a digital twin model and a data-driven neural network intelligent detection model, setting and inputting the digital twin model as a space state to construct a final abnormal detection model;
in this embodiment, the specific process of constructing the final anomaly detection model based on the hybrid algorithm in the ninth step by combining the digital twin model and the data-driven neural network model is as follows:
step 901, setting a digital twin model as a space state and an operation condition of a system-level fusion model, and inputting a fusion algorithm based on internal values of a multi-physics simulation computing system;
step 902, obtaining a preliminary predicted value of each index of the system by a data-driven neural network intelligent detection model obtained by training;
and 903, integrating the digital twin model and the machine learning detection model by using a hybrid Algorithm fusion algorithm to obtain a more accurate detection result, judging whether each index reaches a threshold value, and iteratively updating the abnormal detection model.
Step ten, deploying the trained anomaly detection model on the edge layer to perform real-time anomaly detection; storing the prediction result in an edge layer, verifying the accuracy of the anomaly detection model by comparing the actual running state with the prediction result, evaluating the performance of the model, and determining whether the algorithm needs to be updated or retrained;
step eleven, setting an evaluation mechanism for the output of the abnormality detection model; since anomaly detection is a continuous process, the health of the device is difficult to determine from the values at a particular point in time. The situation that the abnormal condition appears temporarily and the normal range is returned quickly is eliminated, and the condition that the abnormal condition continuously exceeds the normal range is focused; determined by the following evaluation mechanism:
wherein P isi(Yi(t)) is the evaluation result of the detection model output, f is the input signal frequency, when Si(t) the system will make an early warning if it is less than a threshold set based on expertise and experience.
And step twelve, executing a corresponding maintenance plan according to the obtained abnormal detection result, and saving corresponding abnormal data to a database as data for updating the optimization model.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing descriptions of specific exemplary embodiments of the present invention have been presented for purposes of illustration and description. It is not intended to limit the invention to the precise form disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiments were chosen and described in order to explain certain principles of the invention and its practical application to enable one skilled in the art to make and use various exemplary embodiments of the invention and various alternatives and modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims and their equivalents.
Claims (6)
1. A digital twin driven multi-model fusion industrial system anomaly detection method is characterized by comprising the following steps:
firstly, constructing a physical model for a subsystem level and equipment based on a digital twin technology and combining a plurality of software, and integrating the physical model into a unified multi-field system digital twin model;
performing physical model simulation and performing iterative modification and simulation on the multi-field system digital twin model;
thirdly, performing information modeling on the relation between the physical entity and the digital twin body at an edge layer by using an edge computing technology;
collecting data generated by various intelligent sensors in the operation process of the industrial system, and constructing a data-driven neural network intelligent detection model based on past abnormal data of the industrial system;
taking the data obtained in the fourth step as a sample, performing noise reduction, instance selection, normalization, data cleaning and preprocessing on the data obtained in the fourth step on an edge layer by using edge calculation, and processing the problem of sample imbalance by using an XGboost algorithm;
introducing a threshold sorting method to predict abnormal features;
step seven, transmitting the data obtained in the step four to a cloud platform, performing model training, iterating and updating parameters, training a plurality of groups of data sets to obtain an anomaly detection model, wherein the anomaly detection model consists of a plurality of prediction models;
step eight, integrating the abnormal detection models of the subsystems and the equipment in the industrial system through a fusion method, and distributing different weights to form a system-level abnormal detection model;
step nine, based on a hybrid Algorithm fusion algorithm, combining a digital twin model and a data-driven neural network intelligent detection model, setting and inputting the digital twin model as a space state to construct a final abnormal detection model;
step ten, deploying the trained anomaly detection model on the edge layer to perform real-time anomaly detection;
step eleven, setting an evaluation mechanism for the output of the abnormality detection model;
and step twelve, executing a corresponding maintenance plan according to the obtained abnormal detection result, and saving corresponding abnormal data to a database as data for updating the optimization model.
2. The digital twin driven multiple model fusion industrial system anomaly detection method according to claim 1, characterized in that: the specific process of the multi-field system digital twin model constructed on the subsystem level and the equipment in the step one is as follows:
step 101, collecting the operation condition and working condition data of an industrial system provided by an intelligent sensor as input parameters of a model;
102, performing modeling simulation on a control system, a material system, transmission equipment, a production device subsystem and equipment by using ANSYS, MWorks, Dymola and SimlationX software;
103, optimizing the simulation model based on multi-field expert knowledge and empirical data of machinery, heat, mechanics and automation, and integrating models of all subsystems and equipment parts into a multi-field system digital twin model.
3. The digital twin driven multiple model fusion industrial system anomaly detection method according to claim 1, characterized in that: the specific process of performing information modeling on the real-time interaction and synchronization between the physical entity and the digital twin body at the edge layer by using the edge computing technology in the third step is as follows:
step 301, establishing a geometric model for a physical entity at a cloud end for geometric description and kinematic description of a physical system;
step 302, constructing a functional model describing behaviors and functions of the digital twin to manage the functions and behaviors of the digital twin;
step 303, modeling the information of the connection between the physical entity and the digital twin at the edge layer.
4. The digital twin driven multiple model fusion industrial system anomaly detection method according to claim 1, characterized in that: the fourth step is that the data-driven neural network intelligent detection model is constructed, and the specific process is as follows:
step 401, collecting past abnormal data of the industrial system as training data of the model, simultaneously collecting data information of the industrial system obtained by the intelligent sensor under different working conditions, transmitting the data information through MQTT and participating in training to improve the generalization capability of the model;
step 402, constructing an abnormal feature extractor based on a deep neural network, and extracting abnormal features layer by layer through forward propagation;
step 403, mapping the data with different probability distributions in each domain to a high-dimensional space through metric learning, reducing the data distribution difference among different domains, and obtaining the data distribution difference loss;
and step 404, classifying the extracted system operation state features by using a classifier containing a full connection layer to obtain an abnormal detection result and a classification loss, and introducing a regularization term to improve the robustness of the model.
5. The digital twin driven multiple model fusion industrial system anomaly detection method according to claim 1, characterized in that: the specific process of transmitting the data to the cloud platform, training the model, iterating and updating the parameters in the seventh step is as follows:
step 701, collecting and transmitting real-time data to an edge layer through an intelligent sensor and data acquisition equipment, preprocessing the real-time data at the edge layer to reduce transmission pressure, training the preprocessed data on a cloud platform, and storing the data transmitted to the cloud platform in a cloud database;
step 702, determining the most feasible anomaly detection model based on historical data and real-time data through an automatic machine learning optimization algorithm and hyper-parameter selection;
and 703, learning new knowledge from the new real-time data by using incremental learning, adjusting an algorithm and updating an anomaly detection model.
6. The digital twin driven multiple model fusion industrial system anomaly detection method according to claim 1, characterized in that: the concrete process of constructing the final anomaly detection model by combining a digital twin model and a data-driven neural network model based on the hybrid Algorithm fusion algorithm in the step nine is as follows:
step 901, setting a digital twin model as a space state and an operation condition of a system-level fusion model, and inputting a fusion algorithm based on internal values of a multi-physics simulation computing system;
step 902, obtaining a preliminary predicted value of each index of the system by a data-driven neural network intelligent detection model obtained by training;
and 903, integrating the digital twin model and the machine learning detection model by using a hybrid Algorithm fusion algorithm to obtain a more accurate detection result, judging whether each index reaches a threshold value, and iteratively updating the abnormal detection model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111320896.8A CN114048600A (en) | 2021-11-09 | 2021-11-09 | Digital twin-driven multi-model fusion industrial system anomaly detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111320896.8A CN114048600A (en) | 2021-11-09 | 2021-11-09 | Digital twin-driven multi-model fusion industrial system anomaly detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114048600A true CN114048600A (en) | 2022-02-15 |
Family
ID=80207759
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111320896.8A Pending CN114048600A (en) | 2021-11-09 | 2021-11-09 | Digital twin-driven multi-model fusion industrial system anomaly detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114048600A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114595773A (en) * | 2022-03-10 | 2022-06-07 | 广东泰云泽科技有限公司 | Multi-source heterogeneous twin data fusion method and system based on factory production process |
| CN114782417A (en) * | 2022-06-16 | 2022-07-22 | 浙江大学 | Real-time detection method for digital twin characteristics of fan based on edge enhanced image segmentation |
| CN115114342A (en) * | 2022-08-26 | 2022-09-27 | 乘木科技(珠海)有限公司 | Digital twin multi-source data anomaly monitoring method and system |
| CN115489125A (en) * | 2022-09-30 | 2022-12-20 | 吉林大学 | Additive manufacturing digital twin system construction method based on cloud service |
| CN115562191A (en) * | 2022-09-26 | 2023-01-03 | 北京能科瑞元数字技术有限公司 | Productivity intermediate station intelligent conjecture analysis method based on industrial digital twin |
| CN116050678A (en) * | 2023-04-03 | 2023-05-02 | 江苏中科云控智能工业装备有限公司 | Die-casting product processing test system and method based on cloud computing |
| CN116132337A (en) * | 2023-04-04 | 2023-05-16 | 深圳行云创新科技有限公司 | Interface flow anomaly detection method based on service grid technology |
| CN116341396A (en) * | 2023-05-30 | 2023-06-27 | 青岛理工大学 | Complex equipment digital twin modeling method based on multi-source data fusion |
| CN116382197A (en) * | 2023-01-18 | 2023-07-04 | 北京图安世纪科技股份有限公司 | Intelligent factory management platform and management method based on digital twinning |
| CN116962083A (en) * | 2023-09-20 | 2023-10-27 | 西南交通大学 | Method, device and equipment for detecting network abnormal behavior and readable storage medium |
| CN116992516A (en) * | 2023-09-27 | 2023-11-03 | 长春财经学院 | Modeling method and system for bionic product manufactured by digital twin driving additive manufacturing |
| CN117148775A (en) * | 2023-10-31 | 2023-12-01 | 中国电建集团山东电力管道工程有限公司 | Remote monitoring method, system, equipment and medium for pipeline production process |
| CN117332359A (en) * | 2023-12-01 | 2024-01-02 | 国网江苏省电力有限公司南通供电分公司 | Power data transmission abnormality detection method and system |
| CN117371337A (en) * | 2023-12-07 | 2024-01-09 | 安徽金海迪尔信息技术有限责任公司 | Water conservancy model construction method and system based on digital twin |
| CN117786585A (en) * | 2024-02-27 | 2024-03-29 | 威海锐恩电子股份有限公司 | Distribution station terminal abnormality detection method and detection system based on machine learning |
-
2021
- 2021-11-09 CN CN202111320896.8A patent/CN114048600A/en active Pending
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114595773A (en) * | 2022-03-10 | 2022-06-07 | 广东泰云泽科技有限公司 | Multi-source heterogeneous twin data fusion method and system based on factory production process |
| CN114782417A (en) * | 2022-06-16 | 2022-07-22 | 浙江大学 | Real-time detection method for digital twin characteristics of fan based on edge enhanced image segmentation |
| CN115114342A (en) * | 2022-08-26 | 2022-09-27 | 乘木科技(珠海)有限公司 | Digital twin multi-source data anomaly monitoring method and system |
| CN115114342B (en) * | 2022-08-26 | 2022-11-15 | 乘木科技(珠海)有限公司 | Digital twin multi-source data anomaly monitoring method and system |
| CN115562191A (en) * | 2022-09-26 | 2023-01-03 | 北京能科瑞元数字技术有限公司 | Productivity intermediate station intelligent conjecture analysis method based on industrial digital twin |
| CN115562191B (en) * | 2022-09-26 | 2024-02-27 | 北京能科瑞元数字技术有限公司 | Industrial digital twin-based intelligent presumption analysis method for productivity center |
| CN115489125A (en) * | 2022-09-30 | 2022-12-20 | 吉林大学 | Additive manufacturing digital twin system construction method based on cloud service |
| CN116382197A (en) * | 2023-01-18 | 2023-07-04 | 北京图安世纪科技股份有限公司 | Intelligent factory management platform and management method based on digital twinning |
| CN116382197B (en) * | 2023-01-18 | 2023-09-15 | 北京图安世纪科技股份有限公司 | Intelligent factory management platform and management method based on digital twinning |
| CN116050678A (en) * | 2023-04-03 | 2023-05-02 | 江苏中科云控智能工业装备有限公司 | Die-casting product processing test system and method based on cloud computing |
| CN116132337B (en) * | 2023-04-04 | 2023-06-13 | 深圳行云创新科技有限公司 | Interface flow anomaly detection method based on service grid technology |
| CN116132337A (en) * | 2023-04-04 | 2023-05-16 | 深圳行云创新科技有限公司 | Interface flow anomaly detection method based on service grid technology |
| CN116341396A (en) * | 2023-05-30 | 2023-06-27 | 青岛理工大学 | Complex equipment digital twin modeling method based on multi-source data fusion |
| CN116341396B (en) * | 2023-05-30 | 2023-08-11 | 青岛理工大学 | Complex equipment digital twin modeling method based on multi-source data fusion |
| CN116962083B (en) * | 2023-09-20 | 2023-12-05 | 西南交通大学 | Method, device and equipment for detecting network abnormal behavior and readable storage medium |
| CN116962083A (en) * | 2023-09-20 | 2023-10-27 | 西南交通大学 | Method, device and equipment for detecting network abnormal behavior and readable storage medium |
| CN116992516A (en) * | 2023-09-27 | 2023-11-03 | 长春财经学院 | Modeling method and system for bionic product manufactured by digital twin driving additive manufacturing |
| CN116992516B (en) * | 2023-09-27 | 2023-12-12 | 长春财经学院 | Modeling method and system for bionic product manufactured by digital twin driving additive manufacturing |
| CN117148775A (en) * | 2023-10-31 | 2023-12-01 | 中国电建集团山东电力管道工程有限公司 | Remote monitoring method, system, equipment and medium for pipeline production process |
| CN117148775B (en) * | 2023-10-31 | 2024-01-23 | 中国电建集团山东电力管道工程有限公司 | Remote monitoring method, system, equipment and medium for pipeline production process |
| CN117332359A (en) * | 2023-12-01 | 2024-01-02 | 国网江苏省电力有限公司南通供电分公司 | Power data transmission abnormality detection method and system |
| CN117332359B (en) * | 2023-12-01 | 2024-02-09 | 国网江苏省电力有限公司南通供电分公司 | Power data transmission abnormality detection method and system |
| CN117371337A (en) * | 2023-12-07 | 2024-01-09 | 安徽金海迪尔信息技术有限责任公司 | Water conservancy model construction method and system based on digital twin |
| CN117371337B (en) * | 2023-12-07 | 2024-03-15 | 安徽金海迪尔信息技术有限责任公司 | Water conservancy model construction method and system based on digital twin |
| CN117786585A (en) * | 2024-02-27 | 2024-03-29 | 威海锐恩电子股份有限公司 | Distribution station terminal abnormality detection method and detection system based on machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114048600A (en) | Digital twin-driven multi-model fusion industrial system anomaly detection method | |
| CN111596604B (en) | Intelligent fault diagnosis and self-healing control system and method for engineering equipment based on digital twinning | |
| Zan et al. | Control chart pattern recognition using the convolutional neural network | |
| CN110757510B (en) | Method and system for predicting remaining life of robot | |
| CN114282443B (en) | Residual service life prediction method based on MLP-LSTM supervised joint model | |
| CN114266278B (en) | Dual-attention network-based equipment residual service life prediction method | |
| CN115422814B (en) | Digital twin-driven closed-loop optimization design method for complex electromechanical product | |
| CN113837000A (en) | Small sample fault diagnosis method based on task sequencing meta-learning | |
| CN117393076B (en) | Intelligent monitoring method and system for heat-resistant epoxy resin production process | |
| CN115859077A (en) | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions | |
| CN110716520B (en) | Flight control servo actuator reliability evaluation modeling method based on multi-source information fusion | |
| Chen et al. | Transfer life prediction of gears by cross-domain health indicator construction and multi-hierarchical long-term memory augmented network | |
| CN117273440A (en) | Engineering construction Internet of things monitoring and managing system and method based on deep learning | |
| CN115809596A (en) | Digital twin fault diagnosis method and device | |
| CN115964907A (en) | Complex system health trend prediction method and system, electronic device and storage medium | |
| CN111160393A (en) | Data-driven modularized modeling method for carrier rocket health assessment model | |
| Bond et al. | A hybrid learning approach to prognostics and health management applied to military ground vehicles using time-series and maintenance event data | |
| CN109800866B (en) | Reliability increase prediction method based on GA-Elman neural network | |
| CN116070126A (en) | Aviation plunger pump oil distribution disc abrasion detection method and system based on countermeasure self-supervision | |
| CN115718880A (en) | Prediction method for degradation stage of complex equipment | |
| CN114239938A (en) | State-based energy digital twin body construction method | |
| Colace et al. | Unsupervised Learning Techniques for Vibration-Based Structural Health Monitoring Systems Driven by Data: A General Overview | |
| Wang et al. | Complex equipment diagnostic reasoning based on neural network algorithm | |
| CN117314370B (en) | Intelligent energy-based data cockpit system and implementation method | |
| CN116595889B (en) | Processing method and system for thin rib uniform distribution structure based on PEEK material |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |